MONITORING NODE SELECTION ALGORITHM FOR
INTRUSION DETECTION IN CONGESTED SENSOR NETWORK
Jaeun Choi, Myungjong Lee, Gisung Kim and Sehun Kim
Dept. of Industrial & System Engineering, Korea Advanced Institute of Science and Technology, 335 Gwahangno
Yuseong-gu, Daejeon 305-701, Republic of Korea
Keywords: Intrusion detection, Congestion control, Wireless sensor network, Reliable sensor network.
Abstract: Since wireless resources are limited, an efficient way of utilizing wireless resources is needed in selecting
IDSs. We propose a monitoring node selection scheme for intrusion detection in congested wireless sensor
network. Network congestion is an important issue in mobile network. The network congestion does not
guarantee a proper detection rate and congested networks should cause an unreliable network. We consider
congested intrusion detection tasks by queuing theory. We confirm that proposed algorithm guarantee QoS
of monitoring tasks and reliable sensor networks.
1 INTRODUCTION
In recent years, as wireless network techniques
develop rapidly, security becomes one of the major
problems in wireless network. Wireless networks are
notably vulnerable to intrusions, as they operate in
open medium and don’t have any centralized
security systems. In contrast to wired networks,
wireless networks need a special method to detect
intrusions.
An Intrusion Detection System (IDS) for
wireless networks is widely employed for security
purpose to detect illegal intrusions. An IDS is a
system that tries to detect intrusions in the network
using statistical behavior models. To detect
intrusions, all nodes should be monitored by IDS
nodes. Only considering security, every node had
better perform intrusion detection. However, this
approach is inefficient in terms of wireless resources
such as network bandwidth and node battery. If all
nodes implement intrusion detection processing, the
wireless resources are inefficiently used and some
nodes are suffered from battery depletion, because a
node acting as an IDS which overhears and analyses
all packets within monitoring range consumes
additional resources. Since wireless network
resources such as battery and bandwidth are limited,
an efficient way of utilizing these wireless resources
is needed in construction IDSs (Kachirski O. and
Guha R., 1999), (Kim H. et al., 2006).
In this paper, we apply an IDS node selection
scheme to a wireless sensor network (WSN). A
wireless sensor network is a wireless network
consisting of distributed sensors to cooperatively
monitor physical or environmental conditions, such
as temperature, sound, vibration, pressure, motion or
pollutants, at different locations (Römer, Kay;
Friedemann Mattern, 2004), (Thomas Haenselmann,
2006). Wireless sensor networks are now used in
many industrial and civilian application areas
(Römer, Kay; Friedemann Mattern, 2004), (Hadim,
Salem; Nader Mohamed, 2006). Wireless sensor
networks have constraints on resources such as
energy, memory, computational speed and
bandwidth (Römer, Kay; Friedemann Mattern, 2004).
Therefore, an efficient IDS selection algorithm is
required in wireless sensor network.
There are two related works to select monitoring
nodes for intrusion detection before, distributed IDS
(DIDS) and lifetime-enhancing monitoring node
selection (LES) (Kachirski O. and Guha R., 1999),
(Kim H. et al., 2006). These schemes efficiently use
wireless resources such as network lifetime or
battery consumption of whole network. However,
previous researches did not consider congestion of
monitored packets in a buffer of IDS. Our work
differs from others by considering congestion of
monitoring tasks. In recent years, data transmission
speed becomes faster and faster. If the data
transmission rate is faster than monitoring rate, the
unmonitored packets increase in a buffer of IDS.
117
Choi J., Lee M., Kim G. and Kim S. (2009).
MONITORING NODE SELECTION ALGORITHM FOR INTRUSION DETECTION IN CONGESTED SENSOR NETWORK.
In Proceedings of the International Conference on Security and Cryptography, pages 117-120
DOI: 10.5220/0002225001170120
Copyright
c
SciTePress
When congestion occurs in a node due to high
packet arrival rates, the IDS cannot afford to monitor
the all arrival packets and the buffer overflow
happens. So the IDS becomes ineffective and
deteriorated. Hence, it is desirable to assign IDS
node to the network properly so that prevent the
denial of IDS service due to the overflow.
In this paper, we propose a monitoring node
selection algorithm for intrusion detection in
congested sensor network. By using queuing theory,
we prevent congested monitoring situations. In
congested system, our proposed scheme has the
longest network lifetime and the smallest total
battery consumption than other existing schemes. By
preventing congested monitoring tasks, our
algorithm guarantees QoS of monitoring tasks and
reliable sensor networks.
2 ALGORITHM
We propose an IDS node selection scheme based on
two requirements. First, to monitor all nodes in the
network, every node should be in the monitoring
coverage of IDS. Second, to prevent the congestion
of monitored packets, we design the IDS nodes
placement scheme considering congested systems.
For satisfying the first requirement, we apply a set
covering problem (SCP) to the IDS nodes selection
scheme.
The SCP is a classical question in computer
science and complexity theory. The SCP selects a
minimum number of sets that contain all elements
and additionally minimizes the cost of the sets.
Therefore, the SCP guarantees that every element is
covered by at least one server at minimal total cost.
To cover all nodes by minimal IDS, we propose a
formulation using the SCP. The formulation of the
IDS node placement scheme using the SCP is as
follows;
1
1
min
.1
{0,1}
n
jj
j
n
ij j
j
j
cx
s
taxiN
x
jN
=
=
(1)
Formulation (1) is a typical SCP formulation.
Binary variable x
j
is one if node j is IDS node, and
zero otherwise. Like figure 1, binary variable a
ij
is
one if node j is in the transmission range of node i,
and zero otherwise. In the typical SCP, c
j
is the cost
which is needed to select server j. In this problem,
every node has same weight. Therefore, we define c
j
of every node as 1. We define the set N as the set of
all nodes in the network.
Figure 1: Transmission range of a wireless node.
Formulation (1) satisfies the first requirement.
Then, we discuss a constraint considering congested
systems. An implicit assumption in traditional SCP
is that each node in the coverage of a server always
receive satisfied service. However, when a server
suffers from congestion by excessive demand, some
users are not able to receive satisfied service in the
real situation. Especially, if an IDS node suffers
from congestion of monitored packets, intrusion
detection efficiency is reduced and battery
consumption of the IDS is high. In order to
guarantee high detection rate and use efficiently
limited wireless resources, considering congested
systems is important. To prevent congestions, any
packet should not stand in waiting line in the buffer
of IDS nodes for a time longer than a given time-
limit (Marianov, V. and Serra, D., 1998). The
constraint which considers congestion is as follows;
[]P waiting time at IDS node j j
τ
α
.
(2)
Constraint (2) makes the total time spent by a
packet at the IDS node shorter than equal to τ with
probability of at least α. The variables τ and α are
predefined time and probability. In order to express
constraint (2) as a numerical formula, we use the
queuing theory (Marianov, V. and Serra, D., 1998).
In this paper, we make an assumption that an packet
arrival rate from node i to j appears according to a
poisson process with intensity f
ij
. Also, we assume
an exponentially distributed monitoring service time,
with an average rate of μ
j
. This is a reasonable
assumption, since IDS systems behave as M/M/1
queuing systems. As we assume a M/M/1 queuing
system, we are able to use the well known results for
a M/M/1 queuing system for each IDS and its
allocated nodes (Marianov, V. and Serra, D., 1998).
Rewriting constraint (2) as a numerical formula, we
get
SECRYPT 2009 - International Conference on Security and Cryptography
118
1
1
ln (1 )
n
ij ij j j
i
f
ax j N
μα
τ
=
+
.
(3)
Adding constraint (3) to formulation (1), we
finally get our proposed formulation as follows;
1
1
1
min
.1
1
,ln(1)
{0,1}
n
jj
j
n
ij j
j
n
ij ij j j
i
jj
j
cx
st a x i N
fax D
where D j N
xjN
μα
τ
=
=
=
= +
.
(4)
3 PERFORMANCE MEASURE
The normal node uses their energy for transmission
and receiving. Additionally, the monitoring node
overhears packets within monitoring range, and
analyzes them to detect intrusions. The battery
consumed by a monitoring node is calculated as:
()( )( )( )
tt t rr r oo o mm m
E
ms b ms b ms b m s b=++ ++ ++ +
(5)
where, s
t
, s
r
, s
o
and s
m
, respectively, represent the
packet sizes for operations : transmission, receiving,
overhearing and monitoring. Factors m and b are
variable and fixed energy costs for each operation
(Feeney LM, Nilsson M. 2001). We will show
numerical simulations using equation (5).
For comparison with other schemes, we analysis
two performance measures; network lifetime and
sum of remain battery. The network lifetime defined
as the duration of time until the first node runs out of
battery. The network time is a meaningful
performance measure in a sense that a node which
has no battery cannot communicate with any other
nodes and it causes unreliable network (Kachirski O.
and Guha R., 1999). The sum of remaining battery
indicates total battery consumption of whole
network. By analysing this measure, we should
know efficiency of utilizing wireless resources
4 SIMULATION RESULTS
In the simulation, we consider a congested wireless
sensor network consisting of 20 nodes by
MATLAB. For the purpose of comparison, we
consider DIDS and LES mentioned above. We
generate certain amount of packets destined for
random nodes with a packet size of 512 bytes. Also,
we assume that all nodes have the same initial
battery. The simulation results are the average of
100 trials.
Figure 2: Network lifetime.
Figure 3: Sum of remain battery.
Figure 2 shows the average network lifetime in
congested systems and figure 3 shows the sum of
remain battery in congested network. In figure 2, the
network lifetime of proposed scheme is longer than
or similar to those of LES and DIDS. The total
battery consumption in proposed scheme is the
lowest among three schemes. By experiments,
proposed scheme has superior performance than
other existing schemes in congested network.
5 CONCLUSIONS
In this paper, we propose an IDS selection algorithm
for intrusion detection in congested sensor network.
In order to detect intrusions, the packets be buffered
in the queue of a monitoring node. If the buffer is
MONITORING NODE SELECTION ALGORITHM FOR INTRUSION DETECTION IN CONGESTED SENSOR
NETWORK
119
full, intrusion detection is not possible and it causes
the decline of detection rate. It means that congested
network cannot guarantee QoS of intrusion detection.
Moreover, the IDS node which is suffered from
congestion consumes more battery than other normal
IDS nodes.
To prevent congested IDS node, we use queuing
theory in set covering problem. Proposed algorithm
does not select IDS node which is suffered from
congestion of monitoring tasks. By experiment, in
congested situation, our proposed scheme has
superior performance than other existing schemes.
Moreover, we confirm that proposed algorithm
guarantee QoS of monitoring tasks and reliable
sensor network.
In the future works, the verification of our
algorithm is still required to use in the real situation.
Moreover, to reduce calculation time, heuristic
algorithms are required.
ACKNOWLEDGEMENTS
“This research was supported by the Ministry of
Knowledge Economy, Korea, under the
ITRC(Information Technology Research Center)
support program supervised by the IITA(Institute of
Information Technology Advancement)” (IITA-
2009-(C1090-0902-0016)).
REFERENCES
Kachirski O. and Guha R., 1999. Effective Intrusion
Detection Using Multiple Sensors in Wireless Ad Hoc
Networks, Proceeding of the international conference
on system sciences, Hawaii, 2003. p.57-64
Kim H., Kim D. and Kim S., Lifetime-enhancing selection
of monitoring nodes for intrusion detection in mobile
ad hoc networks, International Journal of Electronics
and Communications, (AEU) 60, 2006, p.248-250.
Römer, Kay; Friedemann Mattern, The Design Space of
Wireless Sensor Networks, IEEE Wireless
Communications 11 (6), December 2004, p. 54–61
Thomas Haenselmann, Sensornetworks, GFDL Wireless
Sensor Network textbook, 2006
Hadim, Salem; Nader Mohamed, Middleware Challenges
and Approaches for Wireless Sensor Networks, IEEE
Distributed Systems Online 7 (3), 2006
Marianov, V. and Serra, D., Probabilistic Maximal
Covering Location-Allocation Models for Congested
Systems, Journal of Regional Science 38, 1998, p.401-
424
Feeney LM, Nilsson M. Investigating the energy
consumption of a wireless network interface in an ad
hoc networking environment, Proceedings of
INFOCOM, Anchorage, 2001, p.1549-1557
SECRYPT 2009 - International Conference on Security and Cryptography
120